MAHIPAL SINGH RATHORE

Blue Altair | Duration: 1 Year

• Question Answering AI System: Developed a robust Question Answering AI system, specializing in extracting data (tables and text) from PDFs, Docs, and Text files and enables users to receive intelligent answers, Which significantly enhances data accessibility for users.
  Demonstrated proficiency in utilizing tools such as Streamlit, Haystack, and Langchain. Designed and implemented the backend in Python using Flask APIs for the Question Answering AI system. Developed robust APIs to load AI models and facilitated interaction with different tools, ensuring seamless integration.
  Applied expertise in leveraging large language models, including GPT-3, BERT, T5, etc., to optimize generative or extracted answers.
  Implemented end-to-end solutions for efficient storage and retrieval of data in vector databases like Milvus.
  
• ETL Pipelines and Azure Integration: Led the development of ETL pipelines for ingesting diverse flat files from sources like Sharepoint and OneDrive.
  Utilized Azure Data Factory for seamless data orchestration and transfer.
  Employed Azure Function for efficient preprocessing of files, enhancing the overall data processing workflow.
  

Digitate | Duration: 3.5 Years

• Auto AI Platform Development: Played a key role in the development of an Auto AI platform aimed at automating the creation and deployment of machine learning pipelines.
  Engineered pipelines for diverse datasets, including text, images, and statistical data.
  Designed and implemented pipelines for text and structured data, incorporating essential steps such as Exploratory Data Analysis (EDA), Training, and Model Inference.
  Leveraged MLflow to seamlessly manage the end-to-end machine learning lifecycle.
• Streamlined backend development using Python and Flask APIs to facilitate seamless interaction between Java and AI pipelines. Employed RabbitMQ for efficient queue management, ensuring effective communication between Java components and AI pipelines. Orchestrated a Python server to handle requests from Java components, call specific pipelines, and track completion. Managed memory and space utilization for pipelines, queuing calls when necessary resources were unavailable. Integrated Flask APIs for streamlined communication with EDA pipelines and result sharing among components.
  
• Technological Expertise: Utilized a broad spectrum of ML frameworks, including Tensorflow, PyTorch, Transformers, scikit-learn, MXNet, AutoSklearn, T-pot, and Auto-Weka.
  Implemented Explainable AI using SHAP (Shapley Additive Explanations) to provide transparent insights into the model's decision-making process.
  
• Performance Evaluation and Maintenance: Employed drift detection techniques to assess model performance over an extended period.
  Conducted timely model retraining to ensure sustained accuracy and relevancy.
  
• End-to-End Deployment on Azure: Orchestrated the complete deployment of machine learning pipelines on Azure using tools such as Docker and Jenkins.
  
• Specialized Pipelines: Engineered specialized pipelines tailored for Procurement at Tata Chemicals and automated of HR operations.